Analysis of Wildlife Radio-Tracking Data

By Gary C. White and Robert A. Garrott

With the substantial advances in the miniaturization of electronic components, wildlife biologists now routinely monitor the movements of free-ranging animals with radio-tracking devices. This book explicates the many analytical techniques and computer programs available to extract biological information from the radio tracking data.

• Presentation of software programs for solving specific problems
• Design of radio-tracking studies
• Mechanics of data collection
• Estimation of position by triangulation
• Graphic presentation of animal migration, dispersal, fidelity, and association
• Home range estimation, habitat utilization, and estimation of survival rates and population size

• Language: English
• Publisher: Academic Press
• Release date: Dec 2, 2012
• ISBN: 9780080926575

Book preview

90 91 92 93 9 8 7 6 5 4 3 2 1

Preface

Advances in technology, particularly miniaturization of electronic components, have allowed wildlife biologists to remotely monitor free-ranging animals while they pursue their normal movements and activities. Transmitters, each with a unique identifying frequency, are attached to the animals, and signals from these transmitters are received by the biologists to track the animals. Such use of radio-tracking equipment to obtain biological information about animals is known as radio tracking. The design of radio-tracking studies and the analysis of data collected from such studies is the subject of this book.

For nearly 20 years we have been continually involved in studies which have relied primarily on radio-tracking techniques for the collection of data. These studies have involved a variety of North American vertebrates, such as voles, lemmings, foxes, deer, and elk, and have addressed a diversity of topics, such as daily movements, migrations, habitat use, survival, and population estimation. Like most telemetry users, we have struggled with questions about study design, sample sizes, and appropriate analytical procedures. When conducting extensive searches of the published literature to find answers to these questions, we found an abundance of information on field techniques and radio-tracking equipment but relatively little on design and analysis. The information we did find was usually scattered through a wide variety of publications and did not always satisfy our needs.

Many telemetry users have had similar experiences. The potential to extract biological information from data bases developed with radio-tracking information is often never fully realized. We feel this problem is due primarily to the lack of a generalized text which consolidates the many analytical techniques and computer programs available into a single reference. Such references have been written for capture–recapture methods (cf. White et al., 1978; Arnason and Baniuk, 1977) and line transect density estimation methods (Laake et al., 1979; Gates, 1979) and have proven extremely useful. This book is an attempt to produce a similar reference for those using radio-tracking to investigate free-ranging animals.

We have organized the material in this text into a logical sequence of chapters, with each chapter introducing and discussing a specific telemetry topic or analytical problem. The first three chapters deal with designing radio-tracking studies and the mechanics of data collection, and we recommend that all readers, regardless of their research interests and experience, review these chapters. Chapters 4 and 5 discuss the topics of estimating an animal’s position and triangulation and will be of interest to anyone concerned with spatial data. In Chapter 6 we discuss a number of topics dealing with animal movements, such as graphic presentation, migration, dispersal, fidelity, and animal association, which have received little attention with respect to the development of analytical tools. Chapters 7 through 10 are each devoted to discussing a particular analytical problem which has a relatively large body of literature associated with it, such as home range estimation, habitat utilization, and estimating survival rates and population size. In the final chapter (11) we identify topics we feel would be fruitful areas for future research and discuss the need for a portable, comprehensive software package for the analysis of radio-tracking data.

The recent advances in microcomputer technology have provided researchers and biologists the opportunity to obtain powerful machines at a minimum cost. Hence, most scientists and managers employing radio-tracking techniques also have access to a microcomputer, whereas five years ago, access to a mainframe was often unavailable. Therefore, throughout this text we have emphasized software that can be executed on IBM PCs and compatibles or Apple machines that have DOS software compatibility. Because of the quantitative nature of the material, the reader is expected to have had one or more introductory statistical courses and some skills in basic algebra. We have attempted, as much as possible, to tailor the text to those with such training. Occasionally, however, we have included technical presentations of analytical techniques which require relatively advanced training. We strongly encourage those without a strong quantitative background to skip over these sections.

We are grateful to the following individuals who have provided software for our use and modification: John Cary, James Dunn, Michael Samuel, and Michael Stüwe. Ken Sejkora, Eric Anderson, David Saltz, and Joel Schmutz assisted with workshops that helped develop many of the ideas presented in the text. A. William Alldredge, David R. Anderson, Richard M. Bartmann, I. Lehr Brisbin, Jr., Len H. Carpenter, John Cary, William R. Clark, R. Bruce Gill, Thomas E. Hakonson, and Michael E. Samuel also stimulated discussions that refined our thinking. Members of the FW696 class at Colorado State University and graduate students at the University of Minnesota also contributed to our enthusiasm and illuminated our thinking. I. Lehr Brisbin, Jr. and William R. Clark provided critical reviews of draft manuscripts. Ralph Franklin and D. Heyward Hamilton provided support for our research that kept us motivated to write this book. We are deeply grateful to all these people.

Neither of us could have completed this project without the ardent support of our wives, Liz and Diane. We owe them a debt for relinquishing their time to our demands.

Arnason AN, Baniuk L. User’s Manual: POPAN-2, A Data Maintenance and Analysis System for Recapture Data. Winnipeg, Canada: Department of Computer Science, University of Manitoba; 1977.

Gates CE. In: LINETRAN User’s Guide. College Station, Texas: Institute of Statistics, Texas A&M University; 1979:47.

Laake JL, Burnham KP, Anderson DR. User’s Manual for Program TRANSECT. In: Logan, Utah: Utah State University Press; 1979:26.

White GC, Burnham KP, Otis DL, Anderson DR. User’s Manual for Program CAPTURE. In: Logan, Utah: Utah State University Press; 1978:44.

Chapter 1

Preliminaries

Wildlife radio-tracking data are at least three dimensional, consisting of x and y coordinates in space and a t coordinate in time. Throughout this text an animal’s location will be identified as the vector (x, y, or t). Occasionally, other associated information will be added to this vector, such as elevation above sea level (z), habitat type of the location, height above the ground (i.e., height of a bird in a tree), depth below the water surface (i.e., depth of a dive by a seal), or activity of the animal at the time of the location (feeding, resting, etc.). These attributes augment the basic data obtained by a wildlife tracking study, but often are the main objective of the study. Relating these variables to the x, y, or t vector is an aspect of the data analyses considered in this book.

The topic of this text is the analysis of data collected from tracking free-ranging animals. Our purpose is to provide methods to extract as much information as possible from the data. Before collection of the data begins, some basic decisions must be made as to how to attack the problem. First, a map coordinate system must be chosen for use in the study. Second, a system must be chosen for recording the data for computer processing. The selection of a software system requires that hardware be available to run it. In this chapter we discuss these questions, providing some possibilities to help make these decisions, plus making recommendations based on our experiences.

Map Coordinate Systems

The most familiar of the map coordinate systems is the system of latitude and longitude. This is a circular coordinate system, with any point on the globe delineated on the north–south axis by a latitude coordinate and on the east–west axis by a longitude coordinate. Both coordinates are measured in degrees, minutes, and seconds. The disadvantage of this coordinate system is that it is not rectangular; hence, the calculation of the distance between any two points identified by latitude and longitude is not straightforward. The advantage of this coordinate system with respect to radio-tracking studies is that it is continuous for the entire world. This is important for studies of mobile species which migrate long distances, such as pelagic birds, marine mammals, and many of the bird species which nest in temperate, boreal, and polar regions. Although radio-tracking studies of these species have been severely restricted in the past, due to technological limitations of transmitting and receiving systems, currently emerging satellite technology will undoubtedly result in an increase in investigations of these species. This is probably the best coordinate system to use for such studies which involve very large areas.

A second coordinate system, which is generally less familiar to many telemetry users, is the Universal Transverse Mercator (UTM) system (Edwards 1969, U.S. Army 1973, Snyder 1987), a worldwide rectangular coordinate system used to locate a point on the earth’s surface. Because coordinates on a spherical surface cannot be on a Cartesian coordinate system without breaks or gaps, UTM coordinates are not continuous over the entire globe. Between 80° S latitude and 84° N latitude, the earth has been divided into 60 zones, each zone generally consisting of 6° of longitude (Fig. 1.1). Within each zone the coordinate system is continuous, but orientation of the coordinate systems between zones is slightly different, resulting in breaks in the coordinate system at each zone boundary. Each UTM zone is constructed around the prime meridian for the zone, that is, the longitude in the middle of the zone, with bounding meridians divisible by 6. Zones are numbered from 1 to 60, starting with zone 1 east of 180° W longitude (Fig. 1.1). Thus, the prime meridian for UTM zone 1 is 177° W longitude, and the bounding meridians are 180° to 174° W longitude. The prime meridian for UTM zone 12 is 6° × 12 zones minus 183°, or 111° W longitude, giving the bounding meridians of 114° and 108° W longitude. Note that longitudes west of 0° are negative, that is, 111° W longitude corresponds to – 111° longitude. Because the width of a zone becomes 0 at the North and South poles, the UTM system is generally used only between 80° S latitude and 84° N latitude. From latitudes 84° N and 80° S to their respective poles, the Universal Polar Stereographic (UPS) projection is used instead.

Figure 1.1 A Miller cylindrical projection map of the world illustrating the 60 universal transverse mercator zone designations.

UTM coordinates consist of two 7-digit numbers, with units in meters. The numbers increase as one moves east and north. Within each UTM zone 500,000 m is added to the x coordinates, so that all coordinates are positive (Snyder 1987). Otherwise, all locations west of the prime meridian for the zone would have negative values. The y coordinate indicates the distance (in meters) from the equator, but, due to the zone breaks in the coordinate system, the x coordinate does not provide a similar reference. More or less precise designations are made by adding or subtracting digits. For example, the zone 13 co ordinates (0291923, 4441087) would locate a square meter in northwestern Colorado. However, (0291, 4441) would be less precise and would locate only a square kilometer, because the last three digits of each coordinate have been dropped.

Most radio-tracking studies are regional in nature, involving relatively small areas which usually fall within one UTM zone. The UTM coordinate system is ideal for recording locations in this situation for two reasons. First, it is based on metric measurements, the universal standard for scientists. Second, and more importantly, the UTM system within a zone provides a continuous Cartesian coordinate system, allowing easy calculation of the distances between points and simplifying the calculations used in triangulation, when bearings from different locations are used to estimate an animal’s location. A UTM grid can easily be superimposed over any study area covered by the U. S. Geological Survey 7.5-minute topographical maps. At the edge of each map there are small tick marks labeled with the UTM coordinates. By drawing north-south and east–west lines connecting tick marks with the same coordinates, a 1-km UTM grid is created over the map. Information on the UTM zone and declination of the UTM coordinate system from true north are printed at the lower left margin of the maps (Fig. 1.2).

Figure 1.2 An example of a 1-km UTM coordinate grid drawn on a U. S. Geological Survey 7.5-minute topographical map. (A) Locations of UTM tick marks and their coordinates. (B and C) Locations of information on UTM zone and declination of the coordinate system from true north, respectively. (D) A 100-m grid on drafting film that can be used to obtain more precise locations.

With bad luck, one may end up with a zone change directly across the study area. The quick approach to solving this problem is to extend one of the zones to cover the entire study area. To do this, convert the coordinates from one zone into longitude and latitude, and then convert them into UTM coordinates for the zone of interest. However, if the study area is large, then serious errors can be introduced with this approach. A more complex but less error-prone approach is to select a meridian through the center of the study area as a baseline for a new UTM coordinate system.

A number of computer programs are available that will convert locations from one coordinate system to another. Snyder (1987) provides algorithms for converting between coordinate systems that are commonly used by the U.S. Geological Survey, with numerical examples of conversions provided in an appendix. Tucker and Campbell (1976) provide a set of FORTRAN subroutines for the conversion of locations between coordinate systems. Dodge et al. (1986) provide a program for laptop and personal computers (UTMTEL) that can convert latitude and longitude coordinates to UTM coordinates.

One commonly encountered system that cannot be converted is the township–range land-mapping system used in the United States. Because this system is implemented on the ground through surveys, errors have occurred. Hence, direct conversion to another system is not possible without a database to supply the corrections. Because of the survey errors inherent in the system, distances between locations cannot be reliably computed. Hence, we do not recommend that radio-tracking locations be recorded using the land survey system.

Entry of Data for Computer Processing

Preparing wildlife radio-tracking data for computer analysis requires that the information be entered into a computer database. Many database packages are available to perform this data entry. Before selecting a package, the user must be certain that the package is able to manipulate the data as necessary to perform the analyses described in this text, or at least that the package can easily export the database to a statistical package that can perform the manipulations. Some of the requirements for the database package are that dates and times can be easily entered and manipulated and that error-checking routines can be implemented to check the data as they are entered. The package must be able to handle the volume of data that will be entered; that is, database packages that maintain the entire mass of data in the computer’s memory may not be adequate for a large radio-tracking study. Finally, some simple graphics capabilities are helpful in visually verifying data immediately after entry.

The x, y, and t variables are usually the most important values entered in a computer database. In addition, other variables that should be recorded are the animal’s identity (usually the transmitter’s frequency), age, and sex. Age and sex are often used as covariates or classification variables needed for analyses and do not require much storage capacity. A code which specifies the type of information recorded for the entry should also be included. For example, the place of initial capture should be differentiated from a routine location. The final location (e.g., the site of an animal’s death) should be differentiated for other records. Another example is the last location recorded for an animal before the radio failed. The type of information code would be used to determine whether the records were to be used for home range estimation (routine locations) or survival rate estimation (chronologically, the first and last records).

Entry of the date and the time for computer processing is more complex than entry of the x and y coordinates. Although the Julian date is usually used for analysis, we do not recommend entering data as Julian dates because of the difficulty in visually interpreting values without a ready conversion table. Rather, we suggest that the date and the time be entered as yymmdd.hhmm, where yy is the last two digits of the year, mm is the month (01 – 12), dd is the day of the month (01 – 31), and the decimal point separates the day from the fraction of a day. The hour (hh) is recorded on a 24-hour clock (00–23), and the minute (mm) can have values from 00 to 59. An important reason for recording the date and the time in this format is that the file can be sorted chronologically by time using these 11 columns. The database package must be able to compute from these values the amount of time between two locations.

We advise that the following variables be placed in the file with this order:

An ASCII file in this format can be sorted according to animal identity and time of location using the default SORT command available on most operating systems, although we would use the database system used for data entry for this task. Other information of importance can be entered on the remainder of the record. Specific types of information can be extracted, such as initial capture locations, routine locations of adults, or only locations of deaths of juveniles.

A special problem arises in radio-tracking studies in which the same radio is placed on more than one animal over the course of the study. In such a case a separate identification code might be used for each animal, and two variables are recorded in the file: the animal’s identification code and its transmitter’s frequency. Another complication is also resolved with this convention. An animal can be recaptured and its transmitter replaced with one of a different frequency. Thus, the animal’s identity is not lost when the transmitter frequency changes. The need for this additional variable should be determined in the early stages of the study and the data file constructed accordingly.

In this text, we will emphasize the use of the SAS (SAS Institute Inc. 1985) PC system because of its flexibility and power for programming the various analyses discussed as well as the ability to perform separate analyses by individuals or other group categories. PROC SORT provides a convenient sorting mechanism to reorder the data file. SAS provides the computer functions necessary to convert dates and times into Julian dates and back again for printing in the output. Sorting of the file can be performed using the dates without special handling of the data. Time between two locations can be computed in a variety of units. SAS’s programming language allows computing distances traveled between consecutive locations, or even the creation of a variable containing the total distance traveled during specified intervals. We use SAS because of the availability of high-quality graphics and a variety of statistical analyses. Often, example analyses will be demonstrated in the text with SAS code. SAS provides all of the necessary routine statistical analyses required to summarize wildlife tracking data, including the capability of summarizing the data by individual animals or categories of animals with the BY statement.

In addition, programming analyses are less complicated to do with the SAS DATA step programming language than with a computer language such as FORTRAN or BASIC, primarily because a flexible procedure is provided to input the data and the facilities to sort and manipulate the data are available to the programmer as part of the system. Utilities are also available for both line printers and higher quality output devices, which easily provide visual representations of data. The fundamental reason that we have selected SAS for use in this text over other data analysis systems is the power of the programming language. We are not aware of any other statistical package that provides the potential in its language to program the triangulation methods, home range estimators, or survival analyses presented in this text as SAS programs. SAS is also available on a variety of computers and runs under more than just the DOS operating system generally available on personal computers. Finally, SAS provides an adequate database management system to handle a large radio-tracking project.

Because of the specialized nature of many of the procedures presented, some other specialized programs have become popular, which we discuss as required. However, we do restrict ourselves to software that operates on a personal computer; that is, we do not discuss some of the older packages that are limited to a specific mainframe environment. A brief description of some of the software used in this text appears in Table 1.1. For other packages, the reader will need to acquire the latest version from the software’s author or from an electronic bulletin board. One bulletin board that specializes in wildlife- and fishery-related software is SESAME, located at Raleigh, North Carolina.

Table 1.1

Software Packages of Use in Radio-Tracking Data Analysis on a Personal Computer

Summary

1. Radio-tracking data are three dimensional, consisting of two spatial coordinates (x and y) and one time coordinate (t).

2. The UTM projection system is preferred for representing the spatial coordinates and is assumed throughout the text.

3. All of the analyses presented can be performed on an IBM personal computer or an IBM-compatible computer with the DOS operating system.

4. The structure of the data file to record wildlife radio-tracking data should include at least the animal’s identity, age, and sex; the date and time of the observation; the type of observation; and the UTM coordinates of the animal’s location. Other information might also be required, depending on the objectives of the study.

5. The SAS programming, statistics, and graphics packages are used extensively throughout the text to perform the analyses, and we provide an example code.

6. Additional programs are discussed that perform specific tasks for the analysis of wildlife radio-tracking data.

References

Dodge WE, Wilkie DS, Steiner AJ. UTMTEL: A laptop computer program for location of telemetry finds using LORAN C. Amherst: Massachusetts Cooperative Wildlife Research Unit; 1986 21 pp.

Edwards RL. Archaeological use of the Universal Transverse Mercator grid. Am. Antiquity. 1969;34:180–182.

Heisey DM, Fuller TK. Evaluation of survival and cause-specific mortality rates using telemetry data. J. Wildl. Manage. 1985;49:668–674.

SAS Institute Inc. SAS® Language Guide for Personal Computers, Version 6 Edition. Cary, NC: SAS Institute Inc.; 1985 429 pp.

Snyder JP. In: Map projections—a working manual, Prof. Pap. 1395. U.S. Washington, D.C: Geological Survey; 1987:383.

Tucker TC, Campbell LJ. CATCH: Computer assisted topography, cartography and hypsography. In: Part 2. MAPPROJ: A subroutine package for a number of common map projections, ORNL/TM-3790. Oak Ridge, TN: Oak Ridge Natl. Lab; 1976.

U.S. Army. In: Universal Transverse Mercator Grid, TM 5-241-8. Headquarters. Washington, D.C: Department of the Army; 1973:64.

White GC. Numerical estimation of survival rates from band recovery and biotelemetry data. J. Wildl. Manage.